Dataset used in ExaGAN example

[arun-s-p]

Hi
I am trying to test the ExaGAN application example packed with the lbann repo. I presume that the respective datasets used for training 2D and 3D models, norm_train200k.npy and norm_1_train_val,npy, are sub-sampled from CosmoFlow simulations dataset.
Where can I find details on the sub-sampling process, any normalization and data structure of these files?
Many thanks.

Best regards
Arun